Robots have become required to collaborate with people in real environments. In such collaboration, robots should develop teamwork with people as soon as possible like a capable worker would. Non-verbal behavior plays a crucial role in establishing good team interactions, and it is associated with the influence of facial components such as eyes. However, the necessary facial components of a robot for good human-robot collaboration have not been determined yet. Many robots are designed based on a designer's ideas. Thus, this study focused on blink behavior and an investigation on improving teamwork To tackle this issue, we used one idea of cognitive science: "Joint action" where interacting agents share what is on their minds. We conducted a Go/Nogo test with 29 participants to determine whether or not interacting participants are sharing their minds with the robot as part of a team. In this study, action space was used as an action representing their minds. The participants performed a task with robots having four eye conditions: Blink in-sync, independent blinking, no blinking, and without eyes. The results indicated that a robot that exhibited two blinking behaviors made participants' action space move toward the robot because the area which response times were shorter than far area moved from participants' near area to middle area (center) between the participant and the robot.